Sibling cannibalism in pike, Esox lucius, larvae and juveniles living in outdoor rearing ponds was studied using stomach contents analysis. For the two initial densities tested (6 and 18 larvae m−2, equivalent to 12 and 36 larvae m−3), cannibalism was non-existent during the larval period (13 to 35 mm total length) and was observed only during the juvenile stages. Initial density of larvae influenced both the date of first detection of cannibalistic individuals and the rate of development of cannibalism in the population. At initial stocking densities of 18 larvae m−2 (36 larvae m−3), cannibalism was observed from 21 days after the start of exogenous feeding (mean total length: 60 mm) onwards. At a mean total length of 100 mm and for initial stocking densities of 6 and 18 larvae m−2, (12 and 36 larvae m−3), the average proportions of cannibals in the populations of juveniles were 7.8% and 41.3% and the cannibals accounted for 15.5% and 65.9% of the total pike biomass, respectively. In stomachs of cannibals, young pike were the dominant prey in terms of weight. Dry weights of invertebrate-prey were lower in cannibals than in non-cannibals of similar size. Cannibalism among pike juveniles was characterized by the prey being swallowed whole and head first in the vast majority of cases. There was a strong positive correlation between predator and prey size and the mouth size of a cannibal was found to be an important constraint determining maximum victim size. The overall mean ratio of pike prey length to pike cannibal length was 66.2% and the average ratio of prey head depth to predator mouth width amounted to 87.6%. Prey size selection could be demonstrated for several length-groups of cannibals. These results are compared with the characteristics of early cannibalism in other fish species.